Every once in a while something resonates, and you get the same message from different sources. Like when you suddenly discover a new word that you hadn't known before - but then you see that same new word just a week later!
Last week was a very busy time, as I was flying over to LA last Monday to do the Socal CanJam show this week-end. I like to arrive on the west coast several days in advance so I can try to get over jet lag somewhat, and hopefully do my seminars with some semblance of not sounding half-witted. And anyway it's good to see a few dealers, touch base with Chord and the Sound Organisation, Chord's US distributor.
One of the things I was doing last week was wrapping up some work (for a very high end project) that had taken the summer to code and simulate, and to do lots of listening tests to see what these new modules actually do to the sound quality. One of the aspects concerned filtering out of band noise (think of my DAC's HF filters), and the effect of the Mojo 2 UHD DSP HF filter on this new module. So that aspect was about filtering RF noise and reducing noise floor modulation, something that I recognised as important in the mid 80s.
With all this frenetic activity, I had a Zoom call with Chord and a major recording studio in the UK. They were evaluating Dave, with a view to get a dozen Daves - 3 for each studio. You may ask why 12, as they have 4 studios, so need 3 in each studio. One Dave would be used for the recording process - if a recording engineer wants to use an analogue effect, they need to convert the digital track into analogue to do the effect, and this is where Dave steps in. The second Dave is of course for mastering, so that the engineer can hear the digital recording. The third one is used for the vinyl master cutting lathe - the digital recording needs to be converted to analogue to feed the power amps that drive the cutter head. Chord have already sold a substantial number of Daves into studios and recording engineers around the world. Indeed, a large fraction of modern vinyl releases in the UK actually have Dave in the signal path. Yes I do indeed find that bizarre in the extreme... Buy vinyl with all the attendant distortions, noise and clicks and loss of transparency so that you can hear a Dave?
Anyway, the engineers mentioned the HF filter in Dave - and were surprised as to how substantial the sound quality change was, when there was no significant difference in the measured performance. And that resonated with me as that very morning I had been listening to HF filters and learned something new that had significance that stretched back to the mid 80s and noise floor modulation. Then someone recently posted about a post I made in 2015 about RF noise, noise floor modulation, and my system in the 80's being very variable with the best sound at 2AM with everything turned off in my flat. So I thought that the fates were conspiring to encourage me to talk about this recent experience, as I was reluctant to talk about it as it's a bit technical and somewhat obscure and I was worried that this subject would bore you all.
So go back to the 80's, when by listening to the easily heard effects of mains cables (which in theory should have no difference to sound quality at all) I discovered that RF noise was the reason for the difference. And I proved that assertion by developing a very complex mains filter - 12 LC filters - covering 100kHz to GHz. Once that filter was perfected (long story) I proved that it was RF noise that was the reason for the SQ changes, as mains cables were now completely inaudible before the filter. Once realising that random RF noise was the villain here, it was easy to understand why - random RF noise, when interfering with analogue electronics, would create intermodulation products, and those products would be random, and depend upon the wanted signal level - hence it would create noise floor modulation. No signal, no intermodulation distortion, high signal, high levels of noise from the intermodulation. This is something I have actually measured before, in that using RF filters within DACs reduces measurable noise floor modulation. Most of the time though, it's audible but not measurable.
Once latching onto RF as being vitally important, I then started adding RF filters onto my analogue electronics - pre-amps and power amps. The simplest one is RC filters (RC filter is a simple resistor and capacitor) on the inputs so bandwidth limiting the amps. This you could easily hear - smoother warmer sound quality, with much better instrument separation and focus. And you could play it louder, without it hardening up. This sounded exactly the same as my mains RF filters. And I always ran with as low a filter frequency as I could get away with so that drop at 20k would be less than 1dB, and lowering the frequency always gave a warmer SQ. Now everything added up in my mind - it sounded like RF noise induced noise floor modulation, adding a filter would help reduce RF noise. But I couldn't measure any changes in noise floor modulation. So perhaps in this case, the explanation was not noise floor modulation but the act of bandwidth limiting with RC filters. When you add a simple RC filter, you do three things - reduce the RF noise pick-up, reduce the audio signal HF amplitude (less than 1dB), and add a phase shift. The phase shift gives a delay with high frequencies, but no delay for low frequencies. So perhaps its not noise floor modulation - but the amplitude drop, or the phase shift. But it sounded exactly the same as RF noise filtering on the mains, so if it quacks like duck and looks like a duck... So I was pretty convinced that it was RF induced noise floor modulation, but I couldn't be 100% sure. And that has bothered me for the past 37 years, as I couldn't prove it. And doubt is an essential driver of progress.
But then a few years ago I developed the UHD DSP for Mojo 2. The UHD DSP uses bi-quads, and these are IIR type digital filters, very different to the FIR digital filters that are used in the WTA filters. IIR filters can be made to be identical mathematically to a RC filter, so the same amplitude and phase response as a RC filter. The Mojo 2 UHD DSP had a HF filter built in, but this was more advanced than a first order RC filter - it was actually 4th order, so better HF attenuation, and had a tweak built into the coefficients. This tweak meant that the amplitude response was dead flat to 20kHz. Listening to this filter gave the same sound quality as other bandwidth limiting filters - warmer, smoother sound quality with better instrument separation and focus.
So it sounded exactly like previous analogue RF filters by reducing noise floor modulation - but it categorically ruled out the effect of amplitude drop at 20k being the reason for the change in sound quality. But it still left the possibility that the HF phase change, or the added delay that only occurs at high frequencies (above 20k in this instance) was somehow responsible for the perceived sound quality change.
And that leads us to the listening tests last week. Now one of the modules was an uber HF filter - which I had coded, simulated and listened too during Covid. This is an absolute monster of a filter, and is an FIR filter designed to completely remove HF noise above 20k. And I say completely - one of the things I have discovered over Covid was that there is no effective limit to the sensitivity of the brain to noise floor modulation, and in designing this huge filter I kept going until I could hear no changes - that's why it ended up being a crazy performance filter. Anyway, after doing lots of other listening tests for other things, I thought I would turn on and off the UHD DSP HF filter - the one that is flat in amplitude but has a HF delay. And I ran lots of listening tests, and could hear absolutely no difference between the UHD DSP HF filter on or off. I love it when you can hear no difference whatsoever, as it means you have hit the bottom of the barrel and no further improvements are possible. And the other benefit of a zero difference result is that there is no doubt about the interpretation of the results. So this proved (at least to my satisfaction) that the differences I have been hearing all of these decades with analogue RF filters are noise floor modulation and categorically not amplitude or phase related from the RC filters. A lot of work to prove something I was 99% sure of in the 80's, but I felt a huge amount of pleasure in finally nailing this.
The people that say small differences (errors well below the ear's threshold of hearing) are inaudible are either too lazy to bother to listen objectively, or lack acuity to hear these effects. But they actually miss the point; the problem with listening tests, or subjective evaluation is not being able to hear these things - you don't need double blind listening tests to hear things that proverbially hit you in the face; the problem is not being able to reliably and consistently hear ultra small things, but our interpretation of the sound quality changes. And this is quite apparent with pre and power amp design; many manufactures claim big sound quality benefits by extending the HF response - say 1MHz bandwidth power amps for example. Typically they state that it sounds more transparent with better detail resolution. And of course they get this bandwidth by removing RF filters on the input and lead compensation on the global feedback path. But it's not more transparency but of course more noise floor modulation due to the lack of RF filters. Noise floor modulation makes it sound brighter and more etched - you have a little amount of signal dependent noise spicing up the sound. But it increases listening fatigue, and degrades musicality, or the ability to enjoy music.
So when you are doing listening tests, be very, very careful, it's extremely easy to hear an "improvement" that is actually degrading the musical performance; and don't fall into the trap of expectation bias, just because you think that big expensive component is better doesn't make it so.
Happy listening, Rob
But, I now realise that would be a huge mistake. YOU are digital's best source for great sound from the format. MY only hope for even better sounding digital. So, I've changed my mind - I will never try to convince you that vinyl has something very special to offer, lest I distract you from your digital endeavours. You MUST remain single-mindedly devoted to digital. And so, I will take your vinyl punches on the chin and not be triggered again! 
